We first characterized the cell type-specific Npas4 mRNA expression in the mPFC, a key region associated with stress and reward, using a single-nuclei RNA-sequencing (snRNA-seq) approach. Consistent with the previous reports (Lin et al., 2008; Spiegel et al., 2014), Npas4 is expressed only in the neurons, and we did not detect it in astrocytes or glial cells. Npas4 mRNA is predominantly expressed in excitatory neurons (92.6%) throughout cortical layers 2 and 5/6, while a small fraction (7.4%) were found in multiple classes of GABAergic inhibitory neurons (7.4%), including Adarrb2-, Pvalb-, and Sst-positive neurons (Figure 1A–D). Also, 7% of all mPFC excitatory neurons expressed detectable Npas4 mRNA, opposed to expression in only 2.5% of inhibitory neurons (Figure 1D). Next, we examined the expression of Npas4 mRNA in two key corticolimbic regions, the mPFC and the nucleus accumbens (NAc), following 11 days of CSDS. We compared CSDS responses to a single social defeat stress experience (acute stress; Figure 1E). We observed a very rapid and transient induction of Npas4 mRNA in the mPFC (Figure 1F, two-way ANOVA, F value = 16.6 and df = 77, Tukey’s post hoc analysis: control vs. acute stress at 5 min, p<0.0001, control vs. chronic stress at 5 min, p<0.0001, acute vs. chronic stress at 5 min, p<0.0001, n = 9–10 per group, control vs. acute stress at 15 min, p<0.0001, control vs. chronic stress at 15 min, p<0.0001, acute vs. chronic stress at 15 min, p<0.0001, n = 6–10 per group) and NAc (Figure 1—figure supplement 1A). We observed a similar response with cFos mRNA in the mPFC, albeit a slower induction and longer duration of expression (Figure 1—figure supplement 1B). Interestingly, CSDS-induced expression of both Npas4 and cFos was observed, although it was reduced compared to the acute stress response (Figure 1F, Figure 1—figure supplement 1B), possibly due to CSDS-induced mPFC hypofunction. In contrast, the CSDS-induced attenuation of Npas4 induction was not observed in the NAc (Figure 1—figure supplement 1A). Analogous to stress-induced increases in Npas4 mRNA, we observed a significant increase in NPAS4 protein at 1 hr following CSDS or acute stress exposure in multiple mPFC regions, including the anterior cingulate and prelimbic cortex subregions (Figure 1G, two-way ANOVA, F value = 9.695 and Df = 27, Tukey’s post hoc analysis: control vs. acute stress in anterior cingulate cortex, p=0.0267, control vs. chronic stress in anterior cingulate cortex, p=0.0462, n = 3–5 per group, control vs. acute stress in prelimbic cortex, p=0.0281, control vs. chronic stress in prelimbic cortex, p=0.0474, n = 3–5 per group). Consistent with the snRNA-seq data (Figure 1A–D), the vast majority (>75%) of NPAS4+ neurons in the mPFC were co-expressed with CaMKIIα, a classical protein-marker for excitatory pyramidal neurons (Figure 1H, two-way ANOVA, F value = 5.645 and Df = 24, Tukey’s post hoc analysis: control vs. acute stress in CaMKIIα(+) cells, p=0.0131, control vs. chronic stress in CaMKIIα(+) cells, p<0.0001, n = 8–11 per group), with practically no detectable NPAS4 expression in parvalbumin- or somatostatin-expressing GABAergic interneurons (Figure 1I and J). Similar to Npas4 mRNA, the relative NPAS4 protein expression per cell was highest following acute stress (Figure 1—figure supplement 1C), suggesting that acute stress and CSDS activate a similar number of NPAS4-positive mPFC neurons, but the NPAS4 expression level within each cell is lower following repeated psychosocial stress.

To examine the function of NPAS4 in CSDS-induced behaviors (Figure 2A), we employed a neurotropic AAV-mediated RNA-interference approach to reduce endogenous Npas4 in the mPFC using a prevalidated Npas4 short hairpin RNA (shRNA) (AAV2-Npas4 shRNA^PFC and Figure 2B, paired t-test, t value = 3.7 and Df = 3, p=0.0343, n = 4 per group), which reliably reduces NPAS4 expression in multiple studies, and where knockdown effects have been repeatedly validated using Npas4 conditional KO mice (Lin et al., 2008; Maya-Vetencourt et al., 2012; Taniguchi et al., 2017). Adult male mice (C57BL/6J) received a bilateral injection of AAV2-Npas4 shRNA^PFC or AAV2-shRNA scrambled control (AAV2-SC shRNA^PFC). Mice were subjected to 10 days of CSDS or no stress control condition, and then they were tested for sociability, natural reward preference and motivation, and anxiety-like behavior (Figure 2A). The CSDS-treated SC shRNA^PFC and Npas4 shRNA^PFC mice showed a significant reduction in the time spent interacting with a novel social target, as shown by time spent in the interaction zone in the presence a social target (Figure 2C, SC shRNA^PFC mice, two-way ANOVA, F value = 6.69 and Df = 41, Bonferroni post hoc analysis, interaction partner (-) vs. (+) in control no stress animals, p<0.0001, interaction partner (-) vs. (+) in CSDS animals, p=0.0099, control no stress animals vs. CSDS animals in interaction partner (+), p=0.0005, n = 18–25 per group, Npas4 shRNA^PFC mice, two-way ANOVA, F value = 7.553 and Df = 39, Bonferroni post hoc analysis, interaction partner (-) vs. (+) in control no stress animals, p<0.0001, interaction partner (-) vs. (+) in CSDS animals, p=0.0021, control no stress animals vs. CSDS animals in interaction partner (+), p=0.0034, n = 19–22 per group). In addition, there was a main effect of CSDS, but no significant difference between Npas4 shRNA^PFC vs. SC shRNA^PFC mice in the relative distribution of social interaction ratio in CSDS-treated mice (Figure 2D, two-way ANOVA, main effect of CSDS, F value = 10.01 and Df = 78, p=0.0022, n = 18–25). Both Npas4 shRNA^PFC and SC shRNA^PFC mice showed significantly increased social avoidance time and ratio following CSDS (Figure 2E and F; Figure 2E, SC shRNA^PFC mice, two-way ANOVA, F value = 5.541 and Df = 38, Bonferroni post hoc analysis, control no stress animals vs. CSDS animals in interaction partner (+), p<0.0001, n = 16–24 per group, Npas4 shRNA^PFC mice, two-way ANOVA, F value = 4.666 and Df = 38, Bonferroni post hoc analysis, control no stress animals vs. CSDS animals in interaction partner (+), p=0.0015, n = 19–21 per group; Figure 2F, two-way ANOVA, main effect of CSDS, F value = 15.64 and Df = 76, p=0.0002, n = 16–24 per group), suggesting that mPFC NPAS4 is not required for CSDS-induced social avoidance. However, unlike the CSDS-treated SC shRNA^PFC mice, CSDS-treated Npas4 shRNA^PFC mice did not develop anhedonia-like behavior, as detected by a significant reduction in sucrose preference in the two-bottle choice test (Figure 2G and Figure 2—figure supplement 1A; Figure 2G, two-way ANOVA, F value = 5.548 and Df = 65, Tukey’s post hoc analysis, control no stress vs. CSDS in SC shRNA^PFC mice, p=0.0291, SC shRNA^PFC vs. Npas4 shRNA^PFC mice with CSDS, p=0.0492, n = 11–24 per group). Interestingly, CSDS increased anxiety-like behavior, as measured in the elevated plus maze, in both SC shRNA^PFC and Npas4 shRNA^PFC mice (Figure 2H, two-way ANOVA, main effect of CSDS, F value = 8.087 and Df = 59, p=0.0061, n = 14–18 per group), indicating that mPFC NPAS4 function is required for some, but not all, of the behavioral sequelae of CSDS. These data suggest that the molecular and circuit mechanisms of CSDS-induced social avoidance, anhedonia, and anxiety might be distinct. Moreover, the presence of CSDS-induced social avoidance and anxiety-related behavior in Npas4 shRNA^PFC mice argues against the possibility that they are simply less sensitive to stress and/or have deficits in threat/fear-related learning and memory.

Individuals who suffer from pathological stress often exhibit reduced motivation to pursue rewards (American Psychiatric Association, 2013; Chen et al., 2015); however, it is also commonly reported that a subset of individuals can exhibit positive behavioral outcomes following stress (i.e., stress resilience) (Linley and Joseph, 2004). To examine the role of NPAS4 in CSDS-induced changes in reward motivation, Npas4 shRNA^PFC or SC shRNA^PFC mice were subjected to CSDS or the ‘no stress’ condition, and then they were allowed to self-administer sucrose (sucrose SA) under operant conditions. After stable sucrose SA was established, we examined motivation to work for a sucrose reward using the progressive ratio (PR) schedule of reinforcement. Compared to SC shRNA^PFC controls, Npas4 shRNA^PFC mice displayed no differences in acquisition of sucrose SA (Figure 3A) or operant discrimination learning (nosepokes in the active vs. inactive port) (Figure 3B). Interestingly, Npas4 shRNA^PFC significantly increased PR breakpoint – the maximum number of nose-pokes an animal was willing to perform to receive a single sucrose reward (Figure 3C, two-way ANOVA, main effect of shRNA expression, F value = 5.92 and Df = 58, p=0.0181, n = 13–19 per group), suggesting that reducing levels of mPFC NPAS4 might enhance reward motivation. Notably, in animals susceptible to CSDS, Npas4 shRNA significantly increased motivation to obtain sucrose, with no change in PR breakpoint after Npas4 shRNA in resilient animals (Figure 3D, two-way ANOVA, F value = 5.685 and Df = 31, Bonferroni post hoc analysis, SC shRNA^PFC and Npas4 shRNA^PFC mice in susceptible group, p=0.0353, n = 3–14 per group), suggesting that CSDS-induced mPFC NPAS4 influences natural reward motivation.

CSDS-induced reduction of dendritic spine density on mPFC pyramidal neurons is a putative pathophysiological underpinning of depression-associated behavior (Cerqueira et al., 2007; Colyn et al., 2019; Liston et al., 2006; McKlveen et al., 2013; Ota and Duman, 2013; Qiao et al., 2016; Qu et al., 2018; Shu and Xu, 2017). As such, we quantified dendritic spine density on deep-layer pyramidal neurons in SC shRNA^PFC or Npas4 shRNA^PFC mice after CSDS compared to nonstressed mice. As expected, we observed a CSDS-induced reduction in dendritic spine density in SC shRNA control mice (Figure 4A, top; Figure 4B, left). In contrast, we observed no CSDS-induced changes in mPFC dendritic spine density in Npas4 shRNA^PFC mice (Figure 4A, bottom; Figure 4B, right), suggesting that NPAS4, either directly or indirectly, is required for this chronic stress-induced structural synaptic change in the mPFC (Figure 4B, two-way ANOVA, F value = 9.864 and Df = 162, Tukey’s post hoc analysis, control no stress vs. CSDS in SC shRNA^PFC mice, p=0.0056, SC shRNA^PFC vs. Npas4 shRNA^PFC mice after CSDS, p<0.0001, n = 34–55 dendrites/8 animals per group). Of note, no changes in mPFC dendritic spine density were observed in nonstressed Npas4 shRNA^PFC mice (Figure 4A and B), indicating that steady-state dendritic spine density in adult mPFC pyramidal neurons of nonstressed animals does not require normal NPAS4 expression levels. In addition, neither Npas4 shRNA nor CSDS produced any detectable changes in mean dendritic spine head diameter or distribution (Figure 4—figure supplement 1).

Rodent models show that chronic restraint, unpredictable (Yuen et al., 2012), or social defeat stress (Kuang et al., 2022) decreases excitatory transmission onto mPFC pyramidal neurons. Furthermore, the administration of ketamine increases excitatory transmission in cultured neurons in vitro (Gerhard et al., 2020) and in mPFC pyramidal neurons in vivo (Zhang et al., 2020), and it alleviates symptoms of depression in human MDD patients through increased mPFC activity (Hare and Duman, 2020). In line with these data and our findings on CSDS-induced decreases in dendritic spine density (Figure 4B), CSDS in control animals significantly increased the miniature excitatory postsynaptic current (mEPSC) inter-event interval in the layer 5 mPFC pyramidal neurons of SC shRNA^PFC mice (Figure 4C and D), consistent with a decrease in presynaptic function and/or reduction in synapse number. However, this change in mEPSC frequency was absent in the CSDS-treated Npas4 shRNA^PFC mice (Figure 4C and D; Figure 4C, two-way ANOVA, F value = 14.57, and Df = 4251, Tukey’s post hoc analysis, control no stress vs. CSDS in SC shRNA^PFC mice, p<0.0001, SC shRNA^PFC vs. Npas4 shRNA^PFC mice after CSDS, p<0.0001, n = 648–1240 events/6–12 neurons/2–4 animals per group). Notably, Npas4 shRNA^PFC also produced a significant increase in mEPSC amplitude in mPFC pyramidal neurons (Figure 4E–G; main effect of Npas4 shRNA), but CSDS did not have this effect (Figure 4E–G; Figure 4E, two-way ANOVA, F value = 6.992 and Df = 4301, Tukey’s post hoc analysis, SC shRNA^PFC vs. Npas4 shRNA^PFC mice with control no stress, p<0.0001, SC shRNA^PFC vs. Npas4 shRNA^PFC mice with CSDS, p<0.0001, n = 654–1253 events/6–12 neurons/2–4 animals per group), suggesting that NPAS4 limits glutamatergic synaptic strength on mPFC pyramidal neurons. Finally, CSDS in control animals significantly increased the paired-pulse ratio (PPR) in excitatory pyramidal neurons, suggesting a reduction presynaptic release probability, but this CSDS-induced effect on presynaptic function was blocked by Npas4 shRNA (Figure 4H, Two-way ANOVA, F value = 5.883 and Df = 561, Tukey’s post hoc analysis, control no stress vs. CSDS in SC shRNA^PFC mice, p=0.0002, SC shRNA^PFC vs. Npas4 shRNA^PFC mice with CSDS, p<0.0056, n = 133–191 events/10–17 neurons/3–5 animals per group). Together, our data reveal that NPAS4 in mPFC is required for reductions in excitatory synaptic transmission and synapse density following chronic psychosocial stress, and that NPAS4 limits basal glutamatergic synaptic strength of deep-layer pyramidal neurons.

To analyze the influence of NPAS4 on the mPFC transcriptome, we performed RNA-seq analyses with mPFC tissue isolated from SC shRNA^PFC and Npas4 shRNA^PFC mice. Of the ~700 differentially expressed genes (DEGs, FDR < 0.05, log₂ (FC) > |0.3|) following Npas4 mRNA knockdown in mPFC, 267 were downregulated and 365 were upregulated (Supplementary file 1). Downregulated genes included Spata3, Defb1, Cidea, Psmb10, and Rspo3 and upregulated genes included Arc, Igfn1, Schip1, Apcdd1, and Dapk2 (Figure 5A and B). A subset of these DEGs was independently validated by qRT-PCR using independent mPFC samples isolated from SC shRNA^PFC and Npas4 shRNA^PFC mice 1 hr after acute social defeat or control, no stress conditions, including Npas4;Two-way ANOVA, F value = 6.736 and Df = 24, Tukey’s post hoc analysis, control SC shRNA vs. Npas4 shRNA mice after acute social defeat, p<0.0159, n = 7 animals per group, Ache (acetylcholinesterase; two-way ANOVA, main effect of Npas4 shRNA, F value = 20 and Df = 24, p=0.0002, n = 7 animals per group), Arpp21 (cAMP regulated phosphoprotein 21; two-way ANOVA, main effect of Npas4 shRNA, F value = 7.433 and Df = 24, p=0.0118, n = 7 animals per group), Dhcr7 (7-dehydrocholesterole reductase; two-way ANOVA, main effect of Npas4 shRNA, F value = 10 and Df = 24, p=0.0042, n = 7 animals per group), Hps4 (HPS4 biogenesis of lysosomal organelles complex 3 subunit 2; two-way ANOVA, main effect of Npas4 shRNA, F value = 7.36 and Df = 24, p=0.0121, n = 7 animals per group), Nfix (nuclear factor I X; two-way ANOVA, main effect of Npas4 shRNA, F value = 6.568 and Df = 24, p=0.0171, n = 7 animals per group), and Sst (somatostatin; two-way ANOVA, main effect of Npas4 shRNA, F value = 5.496 and Df = 23, p=0.0281, n = 6–7 animals per group) (Figure 5—figure supplement 1). Interestingly, Npas4 shRNA upregulated DEGs were significantly enriched in the Midnightblue (MB) module of DEGs that was identified by Bagot and colleagues (Figure 5C, top) (Bagot et al., 2016). This module consists of genes that are upregulated in the PFC of resilient mice, and gene ontology (GO) analysis showed significant enrichment of cell–cell signaling and synaptic transmission genes (Bagot et al., 2016). Furthermore, Npas4 shRNA DEGs were significantly enriched in two PsychENCODE modules (Figure 5C, bottom) (Gandal et al., 2018; Wang et al., 2018); Npas4 shRNA-downregulated DEGs showed significant enrichment within gene module M15, an excitatory neuron module of genes that are associated with ribosome function and upregulated in Autism Spectrum Disorder (ASD) and Bipolar Disorder (BD), while Npas4 shRNA-upregulated DEGs showed significant enrichment in gene module M1, an excitatory neuron module of downregulated genes in ASD that are linked to glutamate-driven neuronal excitability (Gandal et al., 2018). Additionally, functional pathway analysis of Npas4 shRNA-downregulated DEGs revealed significant enrichment of genes linked to ribosome function and protein synthesis. Npas4 shRNA-upregulated DEGs showed significant enrichment of glutamatergic synapse-related genes important for synaptic signaling and organization (Figure 5D). To determine whether these mPFC DEGs are putative direct targets of NPAS4, we compared our data to previously published NPAS4 ChIP-seq studies (Brigidi et al., 2019; Kim et al., 2010) and found significant genomic enrichment of NPAS4 binding to promoter, intron, exon, and distal intergenic genomic regions (Figure 5E), suggesting that NPAS4 may directly regulate several key mPFC genes involved in the regulation of glutamatergic synapses and ribosomes. Interestingly, RNA-seq from human postmortem brains (BA8/9) of male MDD patients indicated significant enrichment of differentially expressed genes (p<0.05) in ribosome-related pathways, including 57 significantly upregulated genes (Figure 5F; Labonté et al., 2017). Of note, the enrichment of ribosomal genes was not observed in female MDD brains (BA8/9), where only one gene, RPS28, exhibited significant differential expression. Moreover, the majority (66.2%) of the Npas4 shRNA-downregulated genes from our analysis overlapped with upregulated genes in human MDD patients (Figure 5F), suggesting that Npas4 expression could contribute to vulnerability to depression in the human brain. Finally, NPAS4 ChIP-seq in hippocampal neurons (Brigidi et al., 2019) indicates that NPAS4 directly associates with 55% (42 of 77) of ribosome-related genes classified in the pathway ‘co-translational protein targeting membrane’ (Figure 5G). We detected 92 total ribosome-related genes in the mPFC that are classified in this pathway, with 68 downregulated and 2 upregulated (p<0.05) by Npas4 shRNA (Figure 5G). Finally, we used qRT-PCR to validate several mPFC genes regulated by Npas4 shRNA and acute social defeat stress and found that Npas4 itself was the only regulated transcript at 1 hr following acute social defeat stress (Figure 5—figure supplement 1). Together, our data suggest that mPFC NPAS4 regulates numerous genes related to glutamatergic synapse regulation and ribosomal function and positions it as a key regulator of healthy mPFC function.

For knockdown of endogenous Npas4 mRNA expression in mPFC, a previously validated Npas4 shRNA, specific to the Npas4 gene, or scramble (SC) shRNA control was cloned into the pAAV-shRNA vector as previously described (Lin et al., 2008; Ploski et al., 2011; Ramamoorthi et al., 2011; Taniguchi et al., 2017). The adeno-associated virus serotype 2 (AAV2) vector consists of a CMV promoter driving eGFP with a SV40 polyadenylation signal, followed downstream by a U6 RNA polymerase III promoter and Npas4 shRNA or scrambled (SC) shRNA oligonucleotides, then a polymerase III termination signal – all flanked by AAV2 inverted terminal repeats. AAV2-Npas4 shRNA and SC shRNA were processed for packaging and purification by the UNC Vector Core (Chapel Hill, NC).

C57BL/6 adult male mice were purchased from Jackson Laboratory (ME) and tested between 8 and 20 weeks of age. Mice were allowed access to food and water ad libitum and were kept on a 12 hr light-dark cycle. All procedures were in accordance with Institutional Animal Care and Use (IACUC) guidelines.

Stereotaxic surgery was performed under general anesthesia with a ketamine/xylazine cocktail (120 mg/kg: 16 mg/kg) or isoflurane (induction 4% v/v, maintenance 1–2% v/v). Coordinates to target the mPFC (ventral portion of cingulate, prelimbic, and infralimbic cortices) were +1.85–1.95 mm anterior, +0.75 mm lateral, and 2.65–2.25 mm ventral from bregma (relative to skull) at a 15° angle in all mice (Covington et al., 2010). AAV2-scramble (SC) shRNA (2.9 * 10^⁹ and 1.1 * 10¹² GC/mL) and AAV2-Npas4 shRNA (4.3 * 10^⁹ and 3.1 * 10¹² GC/mL) were delivered using Hamilton syringes or nanoinjectors with pulled glass capillaries at a rate of 0.1 μL/min for 0.4 μL total at the dorsoventral sites, followed by raising the needle and an additional 0.4 μL delivery of virus. After waiting for an additional 5–10 min, needles were completely retracted. Viral placements were confirmed through immunohistochemistry for bicistronic expression of eGFP from the AAV2 viral vectors by experimenters blinded to the experimental conditions. Animals with off-target virus infection or no infection in one or both hemispheres were excluded from the analysis of behavioral phenotypes.

Control C57BL/6J mice (Vgat-cre positive) were live-decapitated at 8 weeks of age. Brains were rapidly extracted into a supplemented 4°C Hibernate A medium, with GlutaMAX supplement (Fisher), B27 supplement (Fisher), and NxGen RNase inhibitor (0.2 U/uL, Lucigen), and incubated for 30 s. The brain was sectioned into 1 mm sections using a brain matrix. The prefrontal cortex (PFC) was dissected with fine forceps, flash frozen on dry ice, and stored at –80°C. The nuclear isolation protocol was modified from Savell et al., 2020. On the day of nuclear dissociation, the frozen tissue was slowly thawed on ice and chopped with a scalpel blade 100 times in two orthogonal directions on a glass Petri dish. The chopped tissue of three PFCs were pooled and transferred to a hypotonic lysis buffer (10 mM Tris-HCl, 10 mM NaCl, 3 mM MgCl2, 0.1% IGEPAL [Sigma] in Nuclease-Free Water). Supplemented Hibernate A medium was added, and the tissue was triturated ~30 times/sample using three glass pipettes of decreasing diameter. The tissue was then filtered using a 40 micron filter (Fisher). The nuclei were isolated using 500 × g centrifugation and washed with 1× PBS + 1.0% BSA and 0.2 U/uL RNase inhibitor. Nuclei were incubated with 7-aminoactinomycin D (7-AAD) (Invitrogen) for 5 min and sorted for 7-AAD-positive single nuclei using fluorescence-activated nuclear sorting on the Aria II. Samples were counted and diluted to 1500 nuclei/uL before immediate processing on the 10X Genomics Single-Cell Protocol by the MUSC Translation Science Lab. Libraries were constructed using the Chromium Single Cell 3’ Library Construction Kit (10X Genomics, v3.1) and sequenced at Vanderbilt’s Next Gen Sequencing Core (Illumina NovaSeq 6000). Raw sequencing data were processed with Cell Ranger (v6.1.2) (PMID: 28091601). Cellranger mkfastq command was used to demultiplex the different samples and cellranger count command was used to generate gene–cell expression matrices. Ambient RNA contamination was inferred and removed using CellBender (v0.232) with standard parameters. Mouse genome mm10 was used for the alignment, and genecode vM25 was used for gene annotation and coordinates (Frankish et al., 2021).

CSDS was performed as previously described (Golden et al., 2011; Krishnan et al., 2007). CD1 retired male breeders (Charles River Laboratory, CA) were single-housed for 3–5 days before CSDS procedures to establish their territorial cage, then pre-screened for aggressive behavior. Experimental C57BL/6J male mice were introduced to the aggressor’s territorial cage, physically contacted and attacked by the aggressor for 5–10 min, and then separated by a clear plastic board with multiple small holes for 24 hr. Experimental mice were introduced to a new CD1 aggressor each day. The no stress control mice were housed with another non-stressed C57BL/6J male mouse, separated by the same plastic board, and the cage partner was changed every day for 10 days of the CSDS experiment.

Brain tissue was collected at the described time point after social defeat stress, or no stress control condition, and kept frozen at –80°C until processed for the following steps. RNA isolation, reverse transcription, and quantitative real-time PCR were carried out as described previously (Taniguchi et al., 2017). Mouse tissue samples were homogenized in QIAzol solution and processed for RNA purification using the miRNeasy kit (QIAGEN, MD) following the manufacture’s protocol. The total RNA was reverse transcribed using SuperScript III (Invitrogen) with a random hexamer primer following the manufacturer’s instructions. Quantitative PCR (qPCR) was performed using SYBR Green (Bio-Rad, CA). The level of mRNA expression was analyzed by the fold change relative to Gapdh or β-tubulin expression. The relative mRNA level was analyzed as the difference from experimental condition relative to controls. Please see Key Resources Table for primer sequences.

Mouse brains were fixed overnight in 4% PFA in 1× PBS and transferred to a 30% sucrose solution in 1× PBS before slicing (40 or 50 μm) with a microtome. The slices were permeabilized and blocked in 3% BSA, 0.3% Triton X-100, 0.2% Tween-20, 3% normal donkey or goat serum in PBS, then incubated with primary antibodies: anti-GFP (1:1000, Aves Labs, Inc, OR; 1:1000–10,000, Invitrogen), anti-NPAS4 (1:1000, rabbit, kindly provided by Dr. Michael Greenberg’s lab), anti-CaMKIIα (1:1000, Enzo, NY, 6G9), anti-somatostatin (SST) (1:1000, Millipore MAB354), and anti-parvalbumin (PV) (1:1000, Millipore MAB1572) in blocking buffer at room temperature for 2–4 hr or at 4°C overnight. Following a series of 1× PBS rinses, slices were incubated for 1–3 hr at room temperature with secondary antibodies (donkey anti-rabbit 488, goat anti-mouse 594, donkey anti-mouse Cy3, or donkey anti-chicken 488) while protected from light. Slices were counterstained with Hoechst, mounted, and coverslipped on glass slides using AquaMount (Thermo Scientific, MA) or ProlongGold (Thermo Scientific, MA) and analyzed with confocal microscopy (Zeiss LSM 880). The expression level of NPAS4 protein in each cell was measured using ImageJ software in CaMKIIα-positive cells under experimenter-blinded conditions.

Social interaction (SI) assay was performed as previously described (Golden et al., 2011; Krishnan et al., 2007). The social interaction assay was performed 24 hr after the last CSDS procedure. The assay was performed in an open-field arena (44 cm × 44 cm) with the social target’s holding cage. The time mice spent in the interaction zone (8 cm from the social target) was examined for 5 min in the absence and then the presence of a novel CD1 mouse, of which the experimental animal never met, under dim red light using AnyMaze 5.1 (Stoelting Co, Wood Dale) or Ethovision 3.0 software (Noldus, Leesburg, VA). The social interaction ratio was calculated as the time spent in the social interaction zone in the presence of an interaction partner divided by the time in the absence.

The sucrose preference procedure was performed as previously described (Taniguchi et al., 2012). Single-housed mice were provided with Division of Laboratory Animal Resources-approved tap water in two identical double-ball-bearing sipper-style bottles for 2 days, followed by 2 days of 1% (w/v) sucrose solution in tap water to allow for acclimation. Mice were then given one bottle containing tap water and another containing the 1% sucrose solution. Consumption from each bottle was measured every 24 hr for 4 days, and bottle positions were swapped each day to avoid potential side bias. The sucrose preference was calculated by percentage of 1% sucrose consumption volume divided by total liquid consumption volume (sucrose + tap water). Measurement of liquid consumption volume was performed with experimenters blinded to conditions.

The elevated plus maze (EPM) was conducted under the bright light (80 Lux at the closed arm) as previously performed (Taniguchi et al., 2017). Mice were positioned in the center of the maze, and behavior was recorded by video tracking using AnyMaze 5.1 or Ethovision 3.0 software as previously performed (Penrod et al., 2019). The time spent in the open arms was recorded for 5 min.

Sucrose self-administration was conducted for 2 hr at the same time each day for 9–10 days of acquisition training, followed by a progressive ratio schedule as described previously (Taniguchi et al., 2017). Briefly, sucrose availability was signaled both by the house light and a light above the active nosepoke hole. Following a poke in the active hole, both availability lights went off and a cue light inside the nose poke hole was illuminated. Sucrose pellets were delivered immediately upon the active nosepoke, followed by a 10 s time-out period. Nose pokes in the inactive hole were without programmed consequences. During a progressive ratio schedule of reinforcement, the requirements for a sucrose delivery were increased on a subsequent sucrose delivery in an exponential manner. Animals were allowed to self-administer until they failed to earn a sucrose pellet in a 60 min time frame. The last sucrose pellet achieved is reported as an indicator of how much animals consumed before reaching breakpoint.

Mouse brains were collected with rapid live decapitation 24 hr after the social interaction assay and fixed overnight in 4% PFA in 1× PB, then transferred to a 30% sucrose solution in 1× PB before slicing (40 μm) with a vibratome. Deep layer eGFP-expressing pyramidal neurons in the prelimbic cortex were sampled for dendritic spine analyses as described previously (Siemsen et al., 2019). Briefly, proximal apical dendrites were imaged with a Leica SP8 laser scanning confocal microscope equipped with HyD detectors for enhanced sensitivity. Dendritic spine segments were selected only if they satisfied the following criteria: (1) could clearly be traced back to a cell body of origin, (2) were not obfuscated by other dendrites, and (3) were proximal to the branch point separating the apical tuft from the proximal apical dendrite. Images were collected with a ×63 oil immersion objective (1.4 N.A.) at 1024 × 1024 frame size, 4.1× digital zoom, and a 0.1 µm Z-step size (0.04 × 0.04 × 0.1 µm voxel size). Pinhole was set at 0.8 airy units and held constant. Laser power and gain were empirically determined and then held relatively constant, only adjusting to avoid saturated voxels. Huygens Software (Scientific Volume Imaging, Hilversum, NL) was used to deconvolve 3D Z-stacks. Deconvolved Z-stacks were then imported into Imaris (version 9.0.1) software (Bitplane, Zurich, CH). The filament tool was then used to trace and assign the dendrite shaft. Dendritic spines were then semi-automatically traced using the autopath function, and an automatic threshold was used to determine dendritic spine head diameter. Variables exported included the average spine head diameter (in µm) as well as the number of dendritic spines per µm of dendrite (spine density). 3–10 segments were sampled per animal, and the average spine head diameter and the spine density were calculated for each segment. Data for each variable was then expressed as number of spine segments/number of animals. All analyses were performed under experimenter-blinded conditions.

All acute-slice electrophysiological experiments were performed in SC and Npas4 shRNA^PFC mice at 12–14 weeks old. Acute coronal slices (300 μm thickness) containing mPFC were prepared in a semi-frozen 300 mOsM dissection solution containing (in mM): 100.0 choline chloride, 2.5 KCl, 1.25 Na₂H₂PO₄, 25.0 NaHCO₃, 25.0 D-glucose, 3.1 Na-pyruvate, 9.0 Na-ascorbate, 7.0 MgCl₂, 0.5 CaCl₂ and 5.0 kynurenic acid (pH 7.4) and was continually equilibrated with 95% O₂ and 5% CO₂ prior to and during the slicing procedure. Slices were transferred to a 315 mOsM normal artificial cerebrospinal fluid (ACSF) solution containing (in mM): 127.0 NaCl, 2.5 KCl, 1.2 Na₂H₂PO4, 24.0 NaHCO₃, 11.0 D-glucose, 1.2 MgCl₂, 2.4 CaCl₂, and 0.4 Na-ascorbate (pH 7.4) to recover at 37°C for 30 min, and then transferred to room temperature ACSF for an additional 30 min prior to recording.

mPFC pyramidal neurons of layer 5 were visualized with infrared differential interference contrast optics (DIC/infrared optics) and identified by their location, apical dendrites, and spiking patterns in response to depolarizing current injection and AAV2-mediated SC shRNA or Npas4 shRNA expression cells were identified by expression of GFP. Unless stated otherwise, all electrophysiological experiments were performed in whole-cell voltage-clamp mode at –70 mV using borosilicate pipettes (4–6 MΩ) made on NARISHIGE puller (NARISHIGE, PG10) from borosilicate tubing (Sutter Instruments) and filled by an internal solution containing (in mM): 140.0 CsMetSO₄, 5.0 KCl, 1.0 MgCl₂, 0.2 EGTA, 11 HEPES, 2 NaATP, 0.2 Na₂GTP (pH 7.2; 290–295 mOsm).

The AMPA-receptor-mediated mEPSCs were recorded in the presence of 100 μM picrotoxin (GABAARs antagonist, Sigma-Aldrich) and TTX (sodium channels blocker, Sigma-Aldrich). Data were recorded in a series of 10 traces (sweeps), 10 s each. At the beginning of each sweep, a depolarizing step (4 mV for 100ms) was generated to monitor series (10–40 MΩ) and input resistance (>400 MΩ). To analyze data, synaptic events were detected via custom parameters in MiniAnalysis software (Synaptosoft, Decatur, GA) and subsequently confirmed by the observer blinded to the experimental conditions. Data were measured until 700 events in a series were analyzed, or until the maximal duration of the series.

Paired EPSC for PPR measurements were generated at –70 mV with the inter-stimulus interval of 50 ms at frequency of 0.05 Hz – 3 stimulus in 1 min. The peak amplitude of the second EPSC (P2) was divided by the peak of the first amplitude (P1) to generate the PPR (P2/P1).

All data (recordings) were acquired and analyzed by amplifier AXOPATCH 200B (Axon Instruments), digitizer BNC2090 (National instruments), and software AxoGraph v1.7.0, Clampfit v8.0 (pClamp, Molecular Devices), and MiniAnalysis Program v6.0.9 (Synaptosoft). Data were filtered at 2 kHz via AXOPATCH 200B amplifier (Axon Instruments) and digitized at 20 kHz via AxoGraph v1.7.0.

Total RNA was isolated from AAV2-mediated eGFP-positive mPFC slices using the QIAGEN RNA purification kit, as described above. Sequencing was performed by BGI genomics using PolyA mRNA isolation, directional RNA-seq library preparation, and a BIGSeq-500 sequencer. Reads were aligned to the mouse mm10 reference genome using STAR (v2.7.1a) (Dobin et al., 2013). Only uniquely mapped reads were retained for further analyses. Quality control metrics were assessed by Picard tool (RRID:SCR_006525) (http://broadinstitute.github.io/picard/). Gencode annotation for mm10 (version M21) was used as reference alignment annotation and downstream quantification. Gene-level expression was calculated using HTseq (v0.9.1) (Anders et al., 2015) using the intersection-strict mode by exon. Counts were calculated based on protein-coding genes from the annotation file.

Counts were normalized using counts per million reads (CPM). Genes with no reads were removed. Differential expression analysis was performed in R using linear modeling as following: lm(gene expression ~ Treatment + Batch). We estimated log2 fold changes and p-values. p-Values were adjusted for multiple comparisons using a Benjamini–Hochberg correction (FDR). Differentially expressed genes were analyzed at FDR <0.05. Mouse Gene IDs were translated into Human Gene IDs using the biomaRt package (v2.46.0) in R (Durinck et al., 2009).

The functional annotation of differentially expressed and co-expressed genes was performed using GOstats (Falcon and Gentleman, 2007). A Benjamini–Hochberg FDR (FDR <0.05) was applied as a multiple comparison adjustment.

Gene set enrichment was performed in R using Fisher’s exact test with the following parameters: alternative = ‘greater,’ confidence level = 0.95. We reported odds ratio (OR) and Benjamini–Hochberg adjusted p-values (FDR).

One-way, two-way, and three-way analyses of variance (ANOVAs) with or without repeated-measures (RM) were used, followed by Bonferroni or Tukey post hoc tests when a significant interaction was revealed, to analyze mRNA expression, number of NPAS4 (+) cells, NPAS4 protein expression in each cell, percentage of CaMKIIα(+) cells, social interaction, social aversion, sucrose preference, elevated plus maze, sucrose self-administration acquisition and discrimination, dendritic spine morphometric data, and breakpoint in the progressive ratio test. All statistics were performed using GraphPad Prism, except SPSS software was used to handle complex datasets (e.g., three-way ANOVAs). Statistical outliers were detected using a Grubbs test and excluded from analysis. All data are presented as the mean ± SEM. Significance was shown as *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, and nonsignificant values were either not noted or shown as n.s.